Gender differences in modifiable risk factors for hip fracture: 10-year follow-up of a prospective study of 0.5 million Chinese adults

Importance Reliable evidence on modifiable risk factors for hip and major osteoporotic fractures in untreated populations could inform prevention strategies for hip fracture worldwide. Objectives To estimate the age- and sex-specific incidence of hip, major osteoporotic and any fractures and examine their associated risk factors in Chinese adults. Design, Setting, and Participants Prospective study of 512,715 adults, aged 30-79 years, recruited for the China Kadoorie Biobank from 10 diverse areas in China in 2004-2008 and followed up for 10 years. Exposures Sociodemographic factors (e.g. age, sex, and education), medical history (e.g. CVD, diabetes, fracture, and rheumatoid arthritis), anthropometry (e.g. weight, waist-hip ratio) and lifestyle factors (e.g. smoking, alcohol drinking, physical activity, diet). Main Outcomes and Measures Age- and sex-specific incidence rates of fracture types, and adjusted hazard ratios (HR) and population attributable fractions (PAF) for individual risk factors associated with fracture types. Results The incidence rates of hip fracture in Chinese adults were 5.1 (95%CI 5.0-5.3) per 10,000 person-years (2616 cases); and were higher in women than men (5.8 [5.5-6.1] vs 4.2 [3.9-4.5]) and increased by about 2 to 3-fold per 10 years older age. Among men, five risk factors for hip fracture including low education (HR=1.23; 95%CI 1.04-1.45), regular smoker (1.22, 1.03-1.45), lower weight (lowest vs other quintiles: 1.59, 1.34-1.88), alcohol drinker (1.18, 1.02-1.36), and prior fracture (1.62, 1.33-1.98) accounted for 44.3% of hip fractures. Among women, five risk factors including lower weight (lowest vs other quintiles: 1.30, 1.15-1.46), low physical activity (lowest vs other quintiles: 1.22, 1.10-1.35), diabetes (1.62, 1.41-1.86), prior fracture (1.54, 1.33-1.77), and self-rated poor health (1.29, 1.13-1.47), accounted for 24.9% of hip fractures. Associations of these risk factors with major osteoporotic (6857 cases) or any (15762 cases) fractures were weaker than those with hip fracture. Conclusions The age- and sex-specific incidence rates of hip fracture in Chinese adults were comparable with those in Western populations. Five potentially modifiable factors accounted for half of all hip fractures in men and one quarter in women.


Study population
The study design was a prospective study. Participants who were aged 30 -79 years at enrolment, were selected for the China Kadoorie Biobank study of the causes of major chronic diseases that account for most deaths in the Chinese population.
Overall, 515 681 people attended the baseline survey between June 2004 and July 2008, of whom 261 (0.05%) withdrew before completion, 2208 (0.4%) were found subsequently to have inadvertently attended the survey twice at different time points and 1 had major data errors. The estimated population response rate was ∼30% (26-38% in the five rural areas and 16-50% in the five urban areas). All individuals who completed a questionnaire and provided consent for participants in the study were included in the present analyses. No participants with a prior history of cardiovascular disease or cancer were excluded. While the China Kadoorie Biobank is not representative of the Chinese population, it collected data from a large number of participants from 5 rural and 5 urban areas that reflected a wide range of exposures and disease outcomes. The study participants did not receive any payment for their participation in the study.

Exposures
The study procedures, involving electronic data collection, ensured that there was virtually no missing data on covariates. While data were collected on the number of pregnancies, age at menarche and age at menopause, but not on use of hormone replacement therapy, these data were not included in the present analyses which focussed on common risk factors available in both men and women. Information was also collected on daily intake of fruit, vegetables, meat, fish and soy< but were not included in the present analyses. Covariates did include rheumatoid arthritis, but not osteoarthritic or non-specific arthritis as these were unrelated to risk of fracture. In the absence of any screening for osteoporosis (using bone imaging) in the study population, it was not possible to use osteoporosis as a risk factor (exposure) in this population. No data were available on the sites of fractures prior to enrolment.

Outcomes
Data on incidence of fracture and osteoporosis involved all hospital admissions were obtained by linkage to national health insurance claims databases and coded using the International Classification of Diseases, 10 th Revision (ICD-10). There were no procedures in place for screening for osteoporosis in this study population. No data were collected on use of medication such as treatments for osteoporosis or chemotherapy for cancer that might alter risks of fracture. The study methodology and procedures for reporting of fracture outcomes were standardized for all 10 study areas. Follow-up for fracture incidence and all other disease outcomes were remarkably complete with <1% missing data. No data were available on the accuracy of reported fractures, but the reporting and adjudication accuracy of total stroke or ischaemic heart disease was over 95% in this study population.

Statistical analyses
Cox Proportional Hazards models were used to estimate the sex-specific hazard ratios (HR) for fracture types associated with individual risk factors, before and after adjustment for other covariates. In univariable analyses, models were stratified by area and adjusted for age. All variables that were significantly associated (P<0.05) with hip fracture in the univariable analyses were subsequently assessed in multivariable analyses after stratifying by area. Cox proportional hazards models were used to assess the shape and strength of the associations between continuous traits (sex-specific quintiles of height, weight, hip and waist circumference, waist-hip ratio, waist-height ratio, BMI and physical activity) and risk of hip fracture risk. For Cox Proportional hazards, the assumption of proportionality was tested and met.
Population attributable fractions (PAFs, expressed as a percentage) assuming a causal relationship were estimated for potentially modifiable risk factors (i.e., excluding age and height), separately in men and women. PAF was calculated using a standard formula: PAF = Pd×(R−1)/R, where Pd was the proportion of fracture cases exposed to the risk factor, and R was the estimated HR associated with each risk factor in the present study. [2][3][4] Additional details about the methodology used to estimate PAFs are provided in eTable 1.
The incidence rates of hip fracture by areas (n=10) were also adjusted for age and levels of five leading (i.e., highest PAF) risk factors associated with hip fracture, using a previously reported formula: where IR was the incidence rate and was the number of events in each area. 5

Sources of Funding
The chief acknowledgment is to the participants, the project staff, staff of the China CDC and its regional offices for access to death and disease registries. Respondents were asked about the frequency of habitual consumption during the previous 12 months and chose among five categories of frequency: daily (9.8%), 4-6 days/week (2.1%), 1-3 days/week (8.6%), monthly (11.1%), never/rarely (68.4%) Non-consumers Consumers 68.4

Medical history and health status
Self-rated poor health Self-rated health status categorized as Excellent, Good, Fair or Poor Poor health (Yes) No (Excellent/Good/Fair) 90

Diabetes
Diabetes was defined as self-reported history of diabetes or screen detected diabetes. Screen-detected diabetes was defined as no self-reported diabetes with a blood glucose level ≥7.0 mmol/l and a fasting time >8 h, a blood glucose level ≥11.1 mmol/l and a fasting time <8 h, or a fasting blood glucose level ≥7.0 mmol/l. Cox regression models stratified by area, and adjust for age (where available) *Includes those with a (self-reported) previous medical diagnosis of diabetes and those detected by blood glucose tests at baseline eFigure 1. Age-adjusted incidence rates of (a) fractures and (c) osteoporosis across 10 areas (b), by sex

Yes
Open circles indicate rural areas and solid circles indicate urban areas. eFigure 7. Associations of risk factors with hip fracture in men and women, after excluding participants with history of fracture, and the first 5 years of follow-up